D1.2 Network Load Balancing

Size: px
Start display at page:

Download "D1.2 Network Load Balancing"

Transcription

1 D1. Network Load Balancing Ronald van der Pol, Freek Dijkstra, Igor Idziejczak, and Mark Meijerink SARA Computing and Networking Services, Science Park 11, 9 XG Amsterdam, The Netherlands June 1 Introduction SARA s contribution to the GigaPort3 network research is in the area of demanding applications. There is still a large gap between the bandwidth that wide area networks can provide and the speed at which a single server can send and receive data on the network. SARA is working on identifying the bottlenecks in the the overall end-to-end performance and trying to remove those bottlenecks in order to get a higher throughput between applications running on different servers. This document describes the network load balancing analysis and tests that SARA has done in order to overcome the limited capacity of a network interface card. The capacity of network cards is and always has been more limited than the capacity of links in the wide area network. The aim in this research is to combine the capacity of multiple network cards in a single server to allow for a single stream that exceeds the capacity of each individual network card. This work builds on server scalability tests previously performed by SARA [1]. The test setup is described in section. Section 3 provides the test results and section summarises the conclusions. Test Setup The tests described in this document were done with two servers. Table 1 lists the specifications of the two servers Server A Server B Brand Dell PowerEdge T Dell Precision WorkStation T3 CPU Intel Xeon Intel Core Duo GHz Memory GiB (3 GiB) MHz GiB ( 1 GiB) MHz Disk." GB 7 RPM HDDs 1." GB 7 RPM HDD." 1 GB Intel X-M SSDs Network Intel GE UTP NICs Intel GE UTP NICs 1 Myricom port GE SFP+ NICs 1 Myricom port GE SFP+ NICs OS Linux..31 ( bit) Linux..31 ( bit) Table 1. Specs of the servers used in the experiments Whilst the servers could each carry up to four Gb/s Ethernet interfaces, only two of these interfaces were used at the same time, due to limitations of the PCI express slots in the servers.

2 In addition to the Gb/s links, the servers were connected to the regular Internet on a 1 Gb/s interface (not shown). This was to ensure that the ssh session to manage the tests would not interfere with the actual test itself. Most experiments were carried out by connecting the Ethernet interfaces on the Myricom NIC to an Arista 71S Ethernet switch, as shown in figure 1. The switch was configured with two separate VLANs, resulting in two logical links between the servers. The links could be used individually, or be grouped together using link bundling, forming one logical Gb/s link. Fig. 1. Experimental setup. The link bundle was configured to deploy a round robin policy, meaning that packets would evenly be distributed over the two links, as opposed to more common policy to distribute the data based on check-summing the source and destination addresses. The round robin policy ensured that even a single stream could exceed the Gb/s capacity of a physical link. The disadvantage of the round robin policy is possible packet reordering because data of a single stream is distributed over two physical links. To study this effect, an additional setup was created so that one of the two links would traverse a 17 ms loopback from Amsterdam to Geneva and back, while the other link would retain its < 1 ms latency. 3 Results Several throughput measurements were done with the topologies described in the previous section. UDP and TCP traffic was sent between the servers with netperf. Each measurement was done with various socket buffer sizes (from 3 kb to 3 MB) and various message lengths (from bytes to 3, bytes). Each run lasted for 3 seconds. 3.1 Single Local GE Link In this test setup both servers were connected to the Ethernet switch with a single GE link. Both ports on the Ethernet switch were in the same VLAN and Ethernet flow control was enabled and in use on both ports. The Ethernet MTU on both servers was set to 9. The topology of this setup is shown in figure. Fig.. Both servers connected with a GE link via the Ethernet switch.

3 The results of sending UDP and TCP traffic with netperf from to are shown in figure 3. buffer size (in kb) buffer size (in kb) Fig. 3. UDP (left) and TCP (right) traffic from to. The x-axis shows the various socket buffer sizes in kilobytes, the y-axis shows various packet sizes in kilobytes and the z-axis shows the throughput in Gbit/s. The graph on the left in figure 3 is the received packet rate on the receiving. The graph on the right of figure 3 shows the TCP throughput. Because TCP is a reliable protocol, the packet rate at both the sending and receiving end are the same. The graph on the left in figure 3 shows that for UDP the link can be saturated for almost all socket and packet sizes. Only when using a packet size of less than about 1 bytes the maximum throughput is not reached. This is also true when the socket buffer size is less than about kilobyte. The graph on the right in figure 3 shows that larger buffer sizes are needed to saturate the link in the case of TCP than is the case when using UDP. The reason is that TCP needs a large enough buffer size at the receiver in order to keep the link filled with packets. When the receiving buffer is too small, the sender needs to wait for an acknowledgement of packets in transit in order to be allowed to send additional packets. The results of sending UDP and TCP traffic the other way around (from to server A) are shown in figure. buffer size (in kb) buffer size (in kb) Fig.. UDP (left) and TCP (right) traffic from to. When comparing the graphs for both traffic directions several observations can be made. The effect of socket buffer size is roughly the same for both directions. However, UDP performs 3

4 slightly better for small packets when sending from to. Server A is the faster of the two servers and therefore it can keep up better with the incoming UDP packets than. For TCP, traffic in both directions performs equally for all packet sizes. 3. Single GE Link with 17 ms latency UDP and TCP traffic was sent over a link with a large roundtrip time. This topology is shown in figure. 17 ms latency Fig.. Both servers connected with a GE link with a 17 ms latency. The results of traffic sent from to are shown in figure. The graph on the left shows the UDP traffic received at and the graph on the right shows the TCP throughput. buffer size (in kb) buffer size (in kb) Fig.. UDP (left) and TCP (right) traffic from to over a link with 17 ms latency. When comparing the results shown in figure with the results shown in figure 3 the following observations can be made. The results for UDP are almost identical. This is to be expected when the packet loss between the short local connection and the longer connection are the same. UDP has no feedback mechanism; packets are just sent from A to B and the distance of the link has no influence. For TCP it is expected that larger socket buffer sizes are needed when the latency increases. This can not be observed in the graph on the right of figure. In theory, an increase in latency from. ms (local connection) to 17 ms would also result in an -fold increase of the socket buffer for reaching the same throughput. It is not clear why this is not reflected in the

5 measurements. Also, the reason for the dip in the TCP throughput for small packet sizes is not clear. 3.3 Ethernet Channel with Two Equal Length GE Links In this test setup both servers were each connected with two GE ports to the Ethernet switch. On the Ethernet switch two VLANs were configured. Each server had one GE port connected to one VLAN and the other GE port connected to the other VLAN, as shown in figure 7. Fig. 7. Experimental setup for bonding of two equal length links. The Ethernet channel was setup between the two servers, not between a server and the Ethernet switch. The Ethernet switch was transparent with respect to the Ethernet channel. Load balancing between the two individual links of the Ethernet channel was done via round robin load balancing. 1 buffer size (in kb) buffer size (in kb) buffer size (in kb) Fig.. Traffic from to over a x GE Ethernet channel (top left: UDP sending rate; top right: UDP receiving rate; bottom: TCP throughput Figure shows the results of sending traffic via the Gbit/s Ethernet channel from to.

6 The top left graph in figure shows the rate of packets sent by, while the top right graph in figure shows the results of the rate of packets received by. Comparing both pictures clearly shows that the throughput measured at the receiving end is much lower for large socket buffer sizes and packet sizes than what is sent by. Indeed it seems that the performance collapses above a certain sender throughput (roughly at 1-13 Gb/s). Other measurements have shown that the packets arrive at the receiving interface, but are dropped somewhere on the way to the netperf application. It is not clear yet where exactly these packets are dropped. Our hypothesis is that the CPU of the receiving server can not keep up, and is dropping packets in the kernel. Since the kernel has no means to signal the sender to slow down, it it flooded, and the performance collapse. Unfortunately, this hypothesis could not yet be confirmed by additional measurements, and further analysis needs to be done. When looking at the packet rate received by, the top right graph in figure, the graph contains a small ridge. Irrespectively of the socket buffer, there is a small ridge around byte IP packet size and a small valley around 9 byte IP packet size. This can possibly be explained by the effect of fragmentation. The Ethernet MTU on the Ethernet channel was set to 9, bytes. The IP packet sizes that were sent during the measurements consisted of a range of increasing sizes of, 17, 9, 9, k, k and 3k bytes. The throughput is increasing for IP packet sizes from to 9 bytes and dropped for IP packet sizes from 9 and higher. 9 bytes plus UDP header does not fit anymore in the 9, byte Ethernet MTU. Therefore, from IP packet size 9 and up, the IP packets are fragmented. Since the load balancing mechanism deploys a round robin policy, this means that 9 of the 9 bytes of each IP packet end up on one link, while the other bytes of each IP packet end up on the other link. This has been verified by other experiments, which showed the same Ethernet frame rate, but radical different bandwidth on each link. Probably, the additional overhead for fragmentation does not outweigh the reduced overhead due to larger packet size for this IP packet sizes which are only marginally larger than the Ethernet segment size. The bottom graph in figure shows the result of sending TCP traffic from to. This TCP throughput is much better than what is achieved with UDP. The traffic rate does not decrease for large socket buffer and packet sizes. In fact, it it around 9 Gbit/s whenever the socket buffer size is larger than about kb. Figure 9 shows the same result but now in the direction of to. When comparing the results presented in figure 9 to those in figure it is clear that the results for traffic from to are much better than the results for traffic from to. Server A is the faster of the two servers. The better results could be explained by the assumption that the packet receiving process on an Ethernet channel needs more resources than the packet sending process on an Ethernet channel. This is probably the case because reassembly must be done at the receiving end. There is still a drop in throughput when sending from to for larger packet sizes. This can be explained by the fact that larger IP packets are fragmented into more Ethernet frames than smaller IP packets. So the Ethernet frame rate is larger when using larger IP packets and thus put more stress on the receiving server which needs to reassemble all those frames into IP packets again. 3. Ethernet Channel with Two GE Links of Different Lengths Figure show the experimental setup where two links to unequal length are bonded. Figures 11 and 1 respectively show the results of the tests from to and from to.

7 buffer size (in kb) buffer size (in kb) buffer size (in kb) Fig. 9. Traffic from to over a x GE Ethernet channel (top left: UDP sending rate; top right: UDP receiving rate; bottom: TCP throughput 17 ms latency Fig.. Experimental setup for bonding of unequal length links. The sending side for the UDP traffic (top left in figures 11 and 1) show a ridge near an IP MTU of 9 bytes, similar to the result with bonding of equal length links (figure 3). As with that test, the ridge can be explained by fragmentation and uneven distribution of data on the two links. Similarly, the results show a clear drop in performance with near % packet loss above a certain throughput of the sender. The performance of TCP has collapsed due to the difference in round trip time. The throughput from to is less then 1 Gb/s, and the throughput from to is negligible. The results were not expected. For small IP packet sizes, there is no fragmentation on the Ethernet layer, and the difference in latency accumulated to packet reordering at the IP layer. For larger IP packet sizes, there is fragmentation, and the difference in latency should already be handled during reassembly of the packets at the Ethernet layer (and would not result in packet reordering on the IP layer). However, there is no difference in the results with and without fragmentation. Instead, the only parameter that influences the performance is the buffer size. 7

8 buffer size (in kilobytes) packet size (in kilobytes) 1 buffer size (in kb) packet size (in kilobytes) buffer size (in kilobytes) 3 1 packet size (in kilobytes).... Fig. 11. Traffic from to over a x GE Ethernet channel (top left: UDP sending rate; top right: UDP receiving rate; bottom: TCP throughput) 1 buffer size (in kb) buffer size (in kb) buffer size (in kb) Fig. 1. Traffic from to over a x GE Ethernet channel (top left: UDP sending rate; top right: UDP receiving rate; bottom: TCP throughput) For TCP, the buffer needs to be at least the round trip time times the throughput. For Gb/s, this would be Gb/s 3 ms = MByte. Since the buffer sized used in this experiment are much smaller than that, it is expected that this causes the poor throughput. This does not explain the collapse in performance for a buffer size larger than kilobytes. Further examination is required to understand the cause of these effects.

9 Conclusions The measurements shown in this document show that saturating a single GE link is possible. However, saturating an Ethernet channel of two GE links proves to be much harder. The measurements done for this document provided a huge amount of information and insight about the various buffers, tuning parameters, hardware implementations, etc. that influence the end-to-end performance. The reason for not being able to reach Gbit/s is not clear. Further investigations are required to examine the exact cause (in the driver, kernel or application) of the packet loss and decreased performance above a certain socket buffer size or packet size. A main reason for limited performance is probably the limited capacity of. Further investigations will be done with faster hardware. References 1. Ronald van der Pol and Freek Dijkstra, D1.1: Server Scalibility https://noc.sara.nl/nrg/publications/ron-d1.1.pdf. Dell PowerEdge T Technical Guidebook https://noc.sara.nl/wiki/images///server-poweredge-t-tech-guidebook.pdf 3. Intel Chipset and Intel Chipset Datasheet datasheet/313.pdf. Intel Ethernet Server and Desktop PCI Express Adapters 9

VMWARE WHITE PAPER 1

VMWARE WHITE PAPER 1 1 VMWARE WHITE PAPER Introduction This paper outlines the considerations that affect network throughput. The paper examines the applications deployed on top of a virtual infrastructure and discusses the

More information

Sockets vs. RDMA Interface over 10-Gigabit Networks: An In-depth Analysis of the Memory Traffic Bottleneck

Sockets vs. RDMA Interface over 10-Gigabit Networks: An In-depth Analysis of the Memory Traffic Bottleneck Sockets vs. RDMA Interface over 1-Gigabit Networks: An In-depth Analysis of the Memory Traffic Bottleneck Pavan Balaji Hemal V. Shah D. K. Panda Network Based Computing Lab Computer Science and Engineering

More information

Network Performance Optimisation and Load Balancing. Wulf Thannhaeuser

Network Performance Optimisation and Load Balancing. Wulf Thannhaeuser Network Performance Optimisation and Load Balancing Wulf Thannhaeuser 1 Network Performance Optimisation 2 Network Optimisation: Where? Fixed latency 4.0 µs Variable latency

More information

Performance Evaluation of VMXNET3 Virtual Network Device VMware vsphere 4 build 164009

Performance Evaluation of VMXNET3 Virtual Network Device VMware vsphere 4 build 164009 Performance Study Performance Evaluation of VMXNET3 Virtual Network Device VMware vsphere 4 build 164009 Introduction With more and more mission critical networking intensive workloads being virtualized

More information

Performance Analysis of IPv4 v/s IPv6 in Virtual Environment Using UBUNTU

Performance Analysis of IPv4 v/s IPv6 in Virtual Environment Using UBUNTU Performance Analysis of IPv4 v/s IPv6 in Virtual Environment Using UBUNTU Savita Shiwani Computer Science,Gyan Vihar University, Rajasthan, India G.N. Purohit AIM & ACT, Banasthali University, Banasthali,

More information

The new frontier of the DATA acquisition using 1 and 10 Gb/s Ethernet links. Filippo Costa on behalf of the ALICE DAQ group

The new frontier of the DATA acquisition using 1 and 10 Gb/s Ethernet links. Filippo Costa on behalf of the ALICE DAQ group The new frontier of the DATA acquisition using 1 and 10 Gb/s Ethernet links Filippo Costa on behalf of the ALICE DAQ group DATE software 2 DATE (ALICE Data Acquisition and Test Environment) ALICE is a

More information

Quantifying the Performance Degradation of IPv6 for TCP in Windows and Linux Networking

Quantifying the Performance Degradation of IPv6 for TCP in Windows and Linux Networking Quantifying the Performance Degradation of IPv6 for TCP in Windows and Linux Networking Burjiz Soorty School of Computing and Mathematical Sciences Auckland University of Technology Auckland, New Zealand

More information

High-Density Network Flow Monitoring

High-Density Network Flow Monitoring Petr Velan petr.velan@cesnet.cz High-Density Network Flow Monitoring IM2015 12 May 2015, Ottawa Motivation What is high-density flow monitoring? Monitor high traffic in as little rack units as possible

More information

Application Note. Windows 2000/XP TCP Tuning for High Bandwidth Networks. mguard smart mguard PCI mguard blade

Application Note. Windows 2000/XP TCP Tuning for High Bandwidth Networks. mguard smart mguard PCI mguard blade Application Note Windows 2000/XP TCP Tuning for High Bandwidth Networks mguard smart mguard PCI mguard blade mguard industrial mguard delta Innominate Security Technologies AG Albert-Einstein-Str. 14 12489

More information

Presentation of Diagnosing performance overheads in the Xen virtual machine environment

Presentation of Diagnosing performance overheads in the Xen virtual machine environment Presentation of Diagnosing performance overheads in the Xen virtual machine environment September 26, 2005 Framework Using to fix the Network Anomaly Xen Network Performance Test Using Outline 1 Introduction

More information

ZEN LOAD BALANCER EE v3.04 DATASHEET The Load Balancing made easy

ZEN LOAD BALANCER EE v3.04 DATASHEET The Load Balancing made easy ZEN LOAD BALANCER EE v3.04 DATASHEET The Load Balancing made easy OVERVIEW The global communication and the continuous growth of services provided through the Internet or local infrastructure require to

More information

TCP Offload Engines. As network interconnect speeds advance to Gigabit. Introduction to

TCP Offload Engines. As network interconnect speeds advance to Gigabit. Introduction to Introduction to TCP Offload Engines By implementing a TCP Offload Engine (TOE) in high-speed computing environments, administrators can help relieve network bottlenecks and improve application performance.

More information

Building High-Performance iscsi SAN Configurations. An Alacritech and McDATA Technical Note

Building High-Performance iscsi SAN Configurations. An Alacritech and McDATA Technical Note Building High-Performance iscsi SAN Configurations An Alacritech and McDATA Technical Note Building High-Performance iscsi SAN Configurations An Alacritech and McDATA Technical Note Internet SCSI (iscsi)

More information

Experiences with MPTCP in an intercontinental multipathed OpenFlow network

Experiences with MPTCP in an intercontinental multipathed OpenFlow network Experiences with MP in an intercontinental multipathed network Ronald van der Pol, Michael Bredel, Artur Barczyk SURFnet Radboudkwartier 273 3511 CK Utrecht, The Netherlands Email: Ronald.vanderPol@SURFnet.nl

More information

Question: 3 When using Application Intelligence, Server Time may be defined as.

Question: 3 When using Application Intelligence, Server Time may be defined as. 1 Network General - 1T6-521 Application Performance Analysis and Troubleshooting Question: 1 One component in an application turn is. A. Server response time B. Network process time C. Application response

More information

Globus Striped GridFTP Framework and Server. Raj Kettimuthu, ANL and U. Chicago

Globus Striped GridFTP Framework and Server. Raj Kettimuthu, ANL and U. Chicago Globus Striped GridFTP Framework and Server Raj Kettimuthu, ANL and U. Chicago Outline Introduction Features Motivation Architecture Globus XIO Experimental Results 3 August 2005 The Ohio State University

More information

I3: Maximizing Packet Capture Performance. Andrew Brown

I3: Maximizing Packet Capture Performance. Andrew Brown I3: Maximizing Packet Capture Performance Andrew Brown Agenda Why do captures drop packets, how can you tell? Software considerations Hardware considerations Potential hardware improvements Test configurations/parameters

More information

ZEN LOAD BALANCER EE v3.02 DATASHEET The Load Balancing made easy

ZEN LOAD BALANCER EE v3.02 DATASHEET The Load Balancing made easy ZEN LOAD BALANCER EE v3.02 DATASHEET The Load Balancing made easy OVERVIEW The global communication and the continuous growth of services provided through the Internet or local infrastructure require to

More information

WAN Transfer Acceleration

WAN Transfer Acceleration WAN Transfer Acceleration Product Description Functionality Interfaces Specifications Index 1 Functionality... 3 2 Integration... 3 3 Interfaces... 4 3.1 Physical Interfaces...5 3.1.1 Ethernet Network...5

More information

HANIC 100G: Hardware accelerator for 100 Gbps network traffic monitoring

HANIC 100G: Hardware accelerator for 100 Gbps network traffic monitoring CESNET Technical Report 2/2014 HANIC 100G: Hardware accelerator for 100 Gbps network traffic monitoring VIKTOR PUš, LUKÁš KEKELY, MARTIN ŠPINLER, VÁCLAV HUMMEL, JAN PALIČKA Received 3. 10. 2014 Abstract

More information

Linux NIC and iscsi Performance over 40GbE

Linux NIC and iscsi Performance over 40GbE Linux NIC and iscsi Performance over 4GbE Chelsio T8-CR vs. Intel Fortville XL71 Executive Summary This paper presents NIC and iscsi performance results comparing Chelsio s T8-CR and Intel s latest XL71

More information

Disk to Network Streaming at 42 Gbit/s

Disk to Network Streaming at 42 Gbit/s Disk to Network Streaming at 42 Gbit/s Ronald van der Pol Outline! " Goal of the project! " 40GE and 100GE Status! " Streaming server! " GLIF demo! " SC10 demo! " Conclusions Goal of the

More information

Microsoft SQL Server 2012 on Cisco UCS with iscsi-based Storage Access in VMware ESX Virtualization Environment: Performance Study

Microsoft SQL Server 2012 on Cisco UCS with iscsi-based Storage Access in VMware ESX Virtualization Environment: Performance Study White Paper Microsoft SQL Server 2012 on Cisco UCS with iscsi-based Storage Access in VMware ESX Virtualization Environment: Performance Study 2012 Cisco and/or its affiliates. All rights reserved. This

More information

Optimizing Network Virtualization in Xen

Optimizing Network Virtualization in Xen Optimizing Network Virtualization in Xen Aravind Menon EPFL, Lausanne aravind.menon@epfl.ch Alan L. Cox Rice University, Houston alc@cs.rice.edu Willy Zwaenepoel EPFL, Lausanne willy.zwaenepoel@epfl.ch

More information

Oracle Database Scalability in VMware ESX VMware ESX 3.5

Oracle Database Scalability in VMware ESX VMware ESX 3.5 Performance Study Oracle Database Scalability in VMware ESX VMware ESX 3.5 Database applications running on individual physical servers represent a large consolidation opportunity. However enterprises

More information

OpenFlow with Intel 82599. Voravit Tanyingyong, Markus Hidell, Peter Sjödin

OpenFlow with Intel 82599. Voravit Tanyingyong, Markus Hidell, Peter Sjödin OpenFlow with Intel 82599 Voravit Tanyingyong, Markus Hidell, Peter Sjödin Outline Background Goal Design Experiment and Evaluation Conclusion OpenFlow SW HW Open up commercial network hardware for experiment

More information

Gigabit Ethernet Design

Gigabit Ethernet Design Gigabit Ethernet Design Laura Jeanne Knapp Network Consultant 1-919-254-8801 laura@lauraknapp.com www.lauraknapp.com Tom Hadley Network Consultant 1-919-301-3052 tmhadley@us.ibm.com HSEdes_ 010 ed and

More information

N5 NETWORKING BEST PRACTICES

N5 NETWORKING BEST PRACTICES N5 NETWORKING BEST PRACTICES Table of Contents Nexgen N5 Networking... 2 Overview of Storage Networking Best Practices... 2 Recommended Switch features for an iscsi Network... 2 Setting up the iscsi Network

More information

Networking Driver Performance and Measurement - e1000 A Case Study

Networking Driver Performance and Measurement - e1000 A Case Study Networking Driver Performance and Measurement - e1000 A Case Study John A. Ronciak Intel Corporation john.ronciak@intel.com Ganesh Venkatesan Intel Corporation ganesh.venkatesan@intel.com Jesse Brandeburg

More information

Comparison of Wired and Wireless HPC Networking Methods

Comparison of Wired and Wireless HPC Networking Methods Comparison of Wired and Wireless HPC Networking Methods Wyatt Gorman wag2@geneseo.edu Table of Contents 1. Experiment Summary 2. Bandwidth 3. Latency 4. Power Usage 5. Scalability 6. Reliability 7. Cost

More information

Where IT perceptions are reality. Test Report. OCe14000 Performance. Featuring Emulex OCe14102 Network Adapters Emulex XE100 Offload Engine

Where IT perceptions are reality. Test Report. OCe14000 Performance. Featuring Emulex OCe14102 Network Adapters Emulex XE100 Offload Engine Where IT perceptions are reality Test Report OCe14000 Performance Featuring Emulex OCe14102 Network Adapters Emulex XE100 Offload Engine Document # TEST2014001 v9, October 2014 Copyright 2014 IT Brand

More information

Building Enterprise-Class Storage Using 40GbE

Building Enterprise-Class Storage Using 40GbE Building Enterprise-Class Storage Using 40GbE Unified Storage Hardware Solution using T5 Executive Summary This white paper focuses on providing benchmarking results that highlight the Chelsio T5 performance

More information

Optimizing Network Virtualization in Xen

Optimizing Network Virtualization in Xen Optimizing Network Virtualization in Xen Aravind Menon EPFL, Switzerland Alan L. Cox Rice university, Houston Willy Zwaenepoel EPFL, Switzerland Abstract In this paper, we propose and evaluate three techniques

More information

Quantifying TCP Performance for IPv6 in Linux- Based Server Operating Systems

Quantifying TCP Performance for IPv6 in Linux- Based Server Operating Systems Cyber Journals: Multidisciplinary Journals in Science and Technology, Journal of Selected Areas in Telecommunications (JSAT), November Edition, 2013 Volume 3, Issue 11 Quantifying TCP Performance for IPv6

More information

Boost Database Performance with the Cisco UCS Storage Accelerator

Boost Database Performance with the Cisco UCS Storage Accelerator Boost Database Performance with the Cisco UCS Storage Accelerator Performance Brief February 213 Highlights Industry-leading Performance and Scalability Offloading full or partial database structures to

More information

Microsoft Exchange Server 2003 Deployment Considerations

Microsoft Exchange Server 2003 Deployment Considerations Microsoft Exchange Server 3 Deployment Considerations for Small and Medium Businesses A Dell PowerEdge server can provide an effective platform for Microsoft Exchange Server 3. A team of Dell engineers

More information

Introduction To Computer Networking

Introduction To Computer Networking Introduction To Computer Networking Alex S. 1 Introduction 1.1 Serial Lines Serial lines are generally the most basic and most common communication medium you can have between computers and/or equipment.

More information

Datacenter Operating Systems

Datacenter Operating Systems Datacenter Operating Systems CSE451 Simon Peter With thanks to Timothy Roscoe (ETH Zurich) Autumn 2015 This Lecture What s a datacenter Why datacenters Types of datacenters Hyperscale datacenters Major

More information

Windows 8 SMB 2.2 File Sharing Performance

Windows 8 SMB 2.2 File Sharing Performance Windows 8 SMB 2.2 File Sharing Performance Abstract This paper provides a preliminary analysis of the performance capabilities of the Server Message Block (SMB) 2.2 file sharing protocol with 10 gigabit

More information

The Bus (PCI and PCI-Express)

The Bus (PCI and PCI-Express) 4 Jan, 2008 The Bus (PCI and PCI-Express) The CPU, memory, disks, and all the other devices in a computer have to be able to communicate and exchange data. The technology that connects them is called the

More information

3. MONITORING AND TESTING THE ETHERNET NETWORK

3. MONITORING AND TESTING THE ETHERNET NETWORK 3. MONITORING AND TESTING THE ETHERNET NETWORK 3.1 Introduction The following parameters are covered by the Ethernet performance metrics: Latency (delay) the amount of time required for a frame to travel

More information

Performance Comparison of Fujitsu PRIMERGY and PRIMEPOWER Servers

Performance Comparison of Fujitsu PRIMERGY and PRIMEPOWER Servers WHITE PAPER FUJITSU PRIMERGY AND PRIMEPOWER SERVERS Performance Comparison of Fujitsu PRIMERGY and PRIMEPOWER Servers CHALLENGE Replace a Fujitsu PRIMEPOWER 2500 partition with a lower cost solution that

More information

PCI Express* Ethernet Networking

PCI Express* Ethernet Networking White Paper Intel PRO Network Adapters Network Performance Network Connectivity Express* Ethernet Networking Express*, a new third-generation input/output (I/O) standard, allows enhanced Ethernet network

More information

Leveraging NIC Technology to Improve Network Performance in VMware vsphere

Leveraging NIC Technology to Improve Network Performance in VMware vsphere Leveraging NIC Technology to Improve Network Performance in VMware vsphere Performance Study TECHNICAL WHITE PAPER Table of Contents Introduction... 3 Hardware Description... 3 List of Features... 4 NetQueue...

More information

Big Data Technologies for Ultra-High-Speed Data Transfer and Processing

Big Data Technologies for Ultra-High-Speed Data Transfer and Processing White Paper Intel Xeon Processor E5 Family Big Data Analytics Cloud Computing Solutions Big Data Technologies for Ultra-High-Speed Data Transfer and Processing Using Technologies from Aspera and Intel

More information

Zeus Traffic Manager VA Performance on vsphere 4

Zeus Traffic Manager VA Performance on vsphere 4 White Paper Zeus Traffic Manager VA Performance on vsphere 4 Zeus. Why wait Contents Introduction... 2 Test Setup... 2 System Under Test... 3 Hardware... 3 Native Software... 3 Virtual Appliance... 3 Benchmarks...

More information

Evaluation Report: Emulex OCe14102 10GbE and OCe14401 40GbE Adapter Comparison with Intel X710 10GbE and XL710 40GbE Adapters

Evaluation Report: Emulex OCe14102 10GbE and OCe14401 40GbE Adapter Comparison with Intel X710 10GbE and XL710 40GbE Adapters Evaluation Report: Emulex OCe14102 10GbE and OCe14401 40GbE Adapter Comparison with Intel X710 10GbE and XL710 40GbE Adapters Evaluation report prepared under contract with Emulex Executive Summary As

More information

Using TrueSpeed VNF to Test TCP Throughput in a Call Center Environment

Using TrueSpeed VNF to Test TCP Throughput in a Call Center Environment Using TrueSpeed VNF to Test TCP Throughput in a Call Center Environment TrueSpeed VNF provides network operators and enterprise users with repeatable, standards-based testing to resolve complaints about

More information

Certification Document macle GmbH GRAFENTHAL R2208 S2 01/04/2016. macle GmbH GRAFENTHAL R2208 S2 Storage system

Certification Document macle GmbH GRAFENTHAL R2208 S2 01/04/2016. macle GmbH GRAFENTHAL R2208 S2 Storage system macle GmbH GRAFENTHAL R2208 S2 Storage system Executive summary After performing all tests, the macle GmbH GRAFENTHAL R2208 S2 has been officially certified according to the Open-E Hardware Certification

More information

Performance Measurement of Wireless LAN Using Open Source

Performance Measurement of Wireless LAN Using Open Source Performance Measurement of Wireless LAN Using Open Source Vipin M Wireless Communication Research Group AU KBC Research Centre http://comm.au-kbc.org/ 1 Overview General Network Why Network Performance

More information

Performance and Recommended Use of AB545A 4-Port Gigabit Ethernet Cards

Performance and Recommended Use of AB545A 4-Port Gigabit Ethernet Cards Performance and Recommended Use of AB545A 4-Port Gigabit Ethernet Cards From Results on an HP rx4640 Server Table of Contents June 2005 Introduction... 3 Recommended Use Based on Performance and Design...

More information

VXLAN Performance Evaluation on VMware vsphere 5.1

VXLAN Performance Evaluation on VMware vsphere 5.1 VXLAN Performance Evaluation on VMware vsphere 5.1 Performance Study TECHNICAL WHITEPAPER Table of Contents Introduction... 3 VXLAN Performance Considerations... 3 Test Configuration... 4 Results... 5

More information

Wireshark in a Multi-Core Environment Using Hardware Acceleration Presenter: Pete Sanders, Napatech Inc. Sharkfest 2009 Stanford University

Wireshark in a Multi-Core Environment Using Hardware Acceleration Presenter: Pete Sanders, Napatech Inc. Sharkfest 2009 Stanford University Wireshark in a Multi-Core Environment Using Hardware Acceleration Presenter: Pete Sanders, Napatech Inc. Sharkfest 2009 Stanford University Napatech - Sharkfest 2009 1 Presentation Overview About Napatech

More information

Iperf Tutorial. Jon Dugan Summer JointTechs 2010, Columbus, OH

Iperf Tutorial. Jon Dugan <jdugan@es.net> Summer JointTechs 2010, Columbus, OH Iperf Tutorial Jon Dugan Summer JointTechs 2010, Columbus, OH Outline What are we measuring? TCP Measurements UDP Measurements Useful tricks Iperf Development What are we measuring? Throughput?

More information

OpenFlow Switching: Data Plane Performance

OpenFlow Switching: Data Plane Performance This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE ICC 21 proceedings OpenFlow : Data Plane Performance Andrea Bianco,

More information

TCP - Introduction. Features of TCP

TCP - Introduction. Features of TCP TCP - Introduction The Internet Protocol (IP) provides unreliable datagram service between hosts The Transmission Control Protocol (TCP) provides reliable data delivery It uses IP for datagram delivery

More information

50. DFN Betriebstagung

50. DFN Betriebstagung 50. DFN Betriebstagung IPS Serial Clustering in 10GbE Environment Tuukka Helander, Stonesoft Germany GmbH Frank Brüggemann, RWTH Aachen Slide 1 Agenda Introduction Stonesoft clustering Firewall parallel

More information

Performance of Host Identity Protocol on Nokia Internet Tablet

Performance of Host Identity Protocol on Nokia Internet Tablet Performance of Host Identity Protocol on Nokia Internet Tablet Andrey Khurri Helsinki Institute for Information Technology HIP Research Group IETF 68 Prague March 23, 2007

More information

10Gb Ethernet: The Foundation for Low-Latency, Real-Time Financial Services Applications and Other, Latency-Sensitive Applications

10Gb Ethernet: The Foundation for Low-Latency, Real-Time Financial Services Applications and Other, Latency-Sensitive Applications 10Gb Ethernet: The Foundation for Low-Latency, Real-Time Financial Services Applications and Other, Latency-Sensitive Applications Testing conducted by Solarflare and Arista Networks reveals single-digit

More information

QoS & Traffic Management

QoS & Traffic Management QoS & Traffic Management Advanced Features for Managing Application Performance and Achieving End-to-End Quality of Service in Data Center and Cloud Computing Environments using Chelsio T4 Adapters Chelsio

More information

ADVANCED NETWORK CONFIGURATION GUIDE

ADVANCED NETWORK CONFIGURATION GUIDE White Paper ADVANCED NETWORK CONFIGURATION GUIDE CONTENTS Introduction 1 Terminology 1 VLAN configuration 2 NIC Bonding configuration 3 Jumbo frame configuration 4 Other I/O high availability options 4

More information

VXLAN: Scaling Data Center Capacity. White Paper

VXLAN: Scaling Data Center Capacity. White Paper VXLAN: Scaling Data Center Capacity White Paper Virtual Extensible LAN (VXLAN) Overview This document provides an overview of how VXLAN works. It also provides criteria to help determine when and where

More information

Virtualization: TCP/IP Performance Management in a Virtualized Environment Orlando Share Session 9308

Virtualization: TCP/IP Performance Management in a Virtualized Environment Orlando Share Session 9308 Virtualization: TCP/IP Performance Management in a Virtualized Environment Orlando Share Session 9308 Laura Knapp WW Business Consultant Laurak@aesclever.com Applied Expert Systems, Inc. 2011 1 Background

More information

VMware vsphere 4.1 Networking Performance

VMware vsphere 4.1 Networking Performance VMware vsphere 4.1 Networking Performance April 2011 PERFORMANCE STUDY Table of Contents Introduction... 3 Executive Summary... 3 Performance Enhancements in vsphere 4.1... 3 Asynchronous Transmits...

More information

IxChariot Virtualization Performance Test Plan

IxChariot Virtualization Performance Test Plan WHITE PAPER IxChariot Virtualization Performance Test Plan Test Methodologies The following test plan gives a brief overview of the trend toward virtualization, and how IxChariot can be used to validate

More information

Network performance in virtual infrastructures

Network performance in virtual infrastructures Network performance in virtual infrastructures A closer look at Amazon EC2 Alexandru-Dorin GIURGIU University of Amsterdam System and Network Engineering Master 03 February 2010 Coordinators: Paola Grosso

More information

High-Speed TCP Performance Characterization under Various Operating Systems

High-Speed TCP Performance Characterization under Various Operating Systems High-Speed TCP Performance Characterization under Various Operating Systems Y. Iwanaga, K. Kumazoe, D. Cavendish, M.Tsuru and Y. Oie Kyushu Institute of Technology 68-4, Kawazu, Iizuka-shi, Fukuoka, 82-852,

More information

Final for ECE374 05/06/13 Solution!!

Final for ECE374 05/06/13 Solution!! 1 Final for ECE374 05/06/13 Solution!! Instructions: Put your name and student number on each sheet of paper! The exam is closed book. You have 90 minutes to complete the exam. Be a smart exam taker -

More information

High-Performance IP Service Node with Layer 4 to 7 Packet Processing Features

High-Performance IP Service Node with Layer 4 to 7 Packet Processing Features UDC 621.395.31:681.3 High-Performance IP Service Node with Layer 4 to 7 Packet Processing Features VTsuneo Katsuyama VAkira Hakata VMasafumi Katoh VAkira Takeyama (Manuscript received February 27, 2001)

More information

Computer Networks CS321

Computer Networks CS321 Computer Networks CS321 Dr. Ramana I.I.T Jodhpur Dr. Ramana ( I.I.T Jodhpur ) Computer Networks CS321 1 / 22 Outline of the Lectures 1 Introduction OSI Reference Model Internet Protocol Performance Metrics

More information

General Pipeline System Setup Information

General Pipeline System Setup Information Product Sheet General Pipeline Information Because of Pipeline s unique network attached architecture it is important to understand each component of a Pipeline system in order to create a system that

More information

Deployment Guide. How to prepare your environment for an OnApp Cloud deployment.

Deployment Guide. How to prepare your environment for an OnApp Cloud deployment. Deployment Guide How to prepare your environment for an OnApp Cloud deployment. Document version 1.07 Document release date 28 th November 2011 document revisions 1 Contents 1. Overview... 3 2. Network

More information

Optimizing SQL Server Storage Performance with the PowerEdge R720

Optimizing SQL Server Storage Performance with the PowerEdge R720 Optimizing SQL Server Storage Performance with the PowerEdge R720 Choosing the best storage solution for optimal database performance Luis Acosta Solutions Performance Analysis Group Joe Noyola Advanced

More information

Insiders View: Network Security Devices

Insiders View: Network Security Devices Insiders View: Network Security Devices Dennis Cox CTO @ BreakingPoint Systems CanSecWest/Core06 Vancouver, April 2006 Who am I? Chief Technology Officer - BreakingPoint Systems Director of Engineering

More information

1000Mbps Ethernet Performance Test Report 2014.4

1000Mbps Ethernet Performance Test Report 2014.4 1000Mbps Ethernet Performance Test Report 2014.4 Test Setup: Test Equipment Used: Lenovo ThinkPad T420 Laptop Intel Core i5-2540m CPU - 2.60 GHz 4GB DDR3 Memory Intel 82579LM Gigabit Ethernet Adapter CentOS

More information

End-System Optimizations for High-Speed TCP

End-System Optimizations for High-Speed TCP End-System Optimizations for High-Speed TCP Jeffrey S. Chase, Andrew J. Gallatin, and Kenneth G. Yocum Department of Computer Science Duke University Durham, NC 27708-0129 chase, grant @cs.duke.edu, gallatin@freebsd.org

More information

LSI MegaRAID CacheCade Performance Evaluation in a Web Server Environment

LSI MegaRAID CacheCade Performance Evaluation in a Web Server Environment LSI MegaRAID CacheCade Performance Evaluation in a Web Server Environment Evaluation report prepared under contract with LSI Corporation Introduction Interest in solid-state storage (SSS) is high, and

More information

Tyche: An efficient Ethernet-based protocol for converged networked storage

Tyche: An efficient Ethernet-based protocol for converged networked storage Tyche: An efficient Ethernet-based protocol for converged networked storage Pilar González-Férez and Angelos Bilas 30 th International Conference on Massive Storage Systems and Technology MSST 2014 June

More information

Analysis of VDI Storage Performance During Bootstorm

Analysis of VDI Storage Performance During Bootstorm Analysis of VDI Storage Performance During Bootstorm Introduction Virtual desktops are gaining popularity as a more cost effective and more easily serviceable solution. The most resource-dependent process

More information

Implementation and Performance Evaluation of M-VIA on AceNIC Gigabit Ethernet Card

Implementation and Performance Evaluation of M-VIA on AceNIC Gigabit Ethernet Card Implementation and Performance Evaluation of M-VIA on AceNIC Gigabit Ethernet Card In-Su Yoon 1, Sang-Hwa Chung 1, Ben Lee 2, and Hyuk-Chul Kwon 1 1 Pusan National University School of Electrical and Computer

More information

Virtualised MikroTik

Virtualised MikroTik Virtualised MikroTik MikroTik in a Virtualised Hardware Environment Speaker: Tom Smyth CTO Wireless Connect Ltd. Event: MUM Krackow Feb 2008 http://wirelessconnect.eu/ Copyright 2008 1 Objectives Understand

More information

Performance Evaluation of Linux Bridge

Performance Evaluation of Linux Bridge Performance Evaluation of Linux Bridge James T. Yu School of Computer Science, Telecommunications, and Information System (CTI) DePaul University ABSTRACT This paper studies a unique network feature, Ethernet

More information

TCP PERFORMANCE IN MOBILE-IP

TCP PERFORMANCE IN MOBILE-IP TCP PERFORMANCE IN MOBILE-IP Foo Chun Choong Department of Electrical Engineering, National University of Singapore ABSTRACT The throughput performance of TCP in Mobile-IP [1] was investigated. Compared

More information

Pump Up Your Network Server Performance with HP- UX

Pump Up Your Network Server Performance with HP- UX Pump Up Your Network Server Performance with HP- UX Paul Comstock Network Performance Architect Hewlett-Packard 2004 Hewlett-Packard Development Company, L.P. The information contained herein is subject

More information

HP Z Turbo Drive PCIe SSD

HP Z Turbo Drive PCIe SSD Performance Evaluation of HP Z Turbo Drive PCIe SSD Powered by Samsung XP941 technology Evaluation Conducted Independently by: Hamid Taghavi Senior Technical Consultant June 2014 Sponsored by: P a g e

More information

TCP/IP: An overview. Syed A. Rizvi

TCP/IP: An overview. Syed A. Rizvi TCP/IP: An overview Syed A. Rizvi TCP/IP The Internet uses TCP/IP protocol suite to establish a connection between two computers. TCP/IP suite includes two protocols (1) Transmission Control Protocol or

More information

Quantum StorNext. Product Brief: Distributed LAN Client

Quantum StorNext. Product Brief: Distributed LAN Client Quantum StorNext Product Brief: Distributed LAN Client NOTICE This product brief may contain proprietary information protected by copyright. Information in this product brief is subject to change without

More information

Research Report: The Arista 7124FX Switch as a High Performance Trade Execution Platform

Research Report: The Arista 7124FX Switch as a High Performance Trade Execution Platform Research Report: The Arista 7124FX Switch as a High Performance Trade Execution Platform Abstract: Many groups are working on reducing trading execution latency - the time from a critical Ethernet frame

More information

Intel DPDK Boosts Server Appliance Performance White Paper

Intel DPDK Boosts Server Appliance Performance White Paper Intel DPDK Boosts Server Appliance Performance Intel DPDK Boosts Server Appliance Performance Introduction As network speeds increase to 40G and above, both in the enterprise and data center, the bottlenecks

More information

PCI Express High Speed Networks. Complete Solution for High Speed Networking

PCI Express High Speed Networks. Complete Solution for High Speed Networking PCI Express High Speed Networks Complete Solution for High Speed Networking Ultra Low Latency Ultra High Throughput Maximizing application performance is a combination of processing, communication, and

More information

Understanding the Performance of an X550 11-User Environment

Understanding the Performance of an X550 11-User Environment Understanding the Performance of an X550 11-User Environment Overview NComputing's desktop virtualization technology enables significantly lower computing costs by letting multiple users share a single

More information

Network Considerations for IP Video

Network Considerations for IP Video Network Considerations for IP Video H.323 is an ITU standard for transmitting voice and video using Internet Protocol (IP). It differs from many other typical IP based applications in that it is a real-time

More information

Accelerating High-Speed Networking with Intel I/O Acceleration Technology

Accelerating High-Speed Networking with Intel I/O Acceleration Technology White Paper Intel I/O Acceleration Technology Accelerating High-Speed Networking with Intel I/O Acceleration Technology The emergence of multi-gigabit Ethernet allows data centers to adapt to the increasing

More information

Measure wireless network performance using testing tool iperf

Measure wireless network performance using testing tool iperf Measure wireless network performance using testing tool iperf By Lisa Phifer, SearchNetworking.com Many companies are upgrading their wireless networks to 802.11n for better throughput, reach, and reliability,

More information

Using Synology SSD Technology to Enhance System Performance Synology Inc.

Using Synology SSD Technology to Enhance System Performance Synology Inc. Using Synology SSD Technology to Enhance System Performance Synology Inc. Synology_SSD_Cache_WP_ 20140512 Table of Contents Chapter 1: Enterprise Challenges and SSD Cache as Solution Enterprise Challenges...

More information

APPENDIX 1 USER LEVEL IMPLEMENTATION OF PPATPAN IN LINUX SYSTEM

APPENDIX 1 USER LEVEL IMPLEMENTATION OF PPATPAN IN LINUX SYSTEM 152 APPENDIX 1 USER LEVEL IMPLEMENTATION OF PPATPAN IN LINUX SYSTEM A1.1 INTRODUCTION PPATPAN is implemented in a test bed with five Linux system arranged in a multihop topology. The system is implemented

More information

White Paper: Performance of Host-based Media Processing

White Paper: Performance of Host-based Media Processing White Paper: Performance of Host-based Media Processing The viability of host-based software in providing media processing capabilities continues to expand the possibilities for voice application development.

More information

How Router Technology Shapes Inter-Cloud Computing Service Architecture for The Future Internet

How Router Technology Shapes Inter-Cloud Computing Service Architecture for The Future Internet How Router Technology Shapes Inter-Cloud Computing Service Architecture for The Future Internet Professor Jiann-Liang Chen Friday, September 23, 2011 Wireless Networks and Evolutional Communications Laboratory

More information

NETWORK LAYER/INTERNET PROTOCOLS

NETWORK LAYER/INTERNET PROTOCOLS CHAPTER 3 NETWORK LAYER/INTERNET PROTOCOLS You will learn about the following in this chapter: IP operation, fields and functions ICMP messages and meanings Fragmentation and reassembly of datagrams IP

More information

Discussion Paper Category 6 vs Category 5e Cabling Systems and Implications for Voice over IP Networks

Discussion Paper Category 6 vs Category 5e Cabling Systems and Implications for Voice over IP Networks Discussion Paper Category 6 vs Category 5e Cabling Systems and Implications for Voice over IP Networks By Galen Udell Belden CDT Networking 2006 Category 6 vs Category 5e Cabling Systems and Implications

More information